Projects
Research prototypes built for human subjects studies, AI systems designed and deployed, and games. The research projects are described in more depth on the research page.
Research — Human-AI Interaction
-
Automation Confusion2023
Grounded theory of how users form and maintain incorrect mental models of AI. Twelve-category taxonomy. CHI 2023.
-
Awareness Cues and Trust2022
Controlled study of intention cues in shared control. Appropriate trust improved without increasing trust frequency.
-
Two Heads Are Better Than One2022
Dimension space for designing shared-control systems. Survey of 55 systems across six domains. CHI 2022.
-
Playing with Dezgo2023
Case study of mental model formation in generative AI. Players infer AI capabilities from task demand, not observation. FDG 2023.
-
Happy Driver2021
Mood and driving style preference in self-driving cars. Mismatch between mood and style predicts dissatisfaction. HAI 2021.
Research — Accessibility
-
Personalizable Partial Automation2019–2021
AI copilots for game accessibility in spinal cord injury rehabilitation. Six participants, all could play. CHI PLAY 2021.
-
Beyond Fun2021
Psychological benefits of rehabilitation gaming for SCI. Autonomy, competence, relatedness. ASSETS 2021.
-
You Should Have Stayed Home2022–2024
Audience interaction design for VR theatre. Design space across participation, agency, and embodiment. CHI PLAY 2022, CHI 2024.
AI Systems
-
Card Czar2024
Computer vision system for Magic card inventory management. CNN classification, data-driven pricing, O(1) retrieval. In production.
-
Flibbertigibbet2024
Voice-controlled game with hybrid LLM and classical planning architecture. Predates LLM tool-calling.
-
Latent Time-Adaptive Drift-Diffusion Model2018
Deep cognitive model of binary decision timing. Learns sparse temporal sequences orders of magnitude faster than LSTM. arXiv 2021.
Games
-
Oui, Chef!!2018
Restaurant management game where players train neural networks. Supervised learning as core mechanic. GI 2019.
-
My Little Paddle2021
Browser-based DQN reinforcement learning on Pong. Implemented from scratch in C# for Unity WebGL.
-
Para Bellum2018
Multiplayer browser-based space shooter. Node.js, MongoDB, Socket.IO real-time state synchronisation.